<Back to home >

Live Demonstrations

A live demonstrations session will be held in order to illustrate some of the supporting projects' activities. This session allows researchers to directly exchanges ideas with the presenters. Moreover, these demos illustrate more in details some of the presented talks, offering additional insights on the projects research. As follows, each demonstration is shortly depicted.

Deformable Model-based Hip Joint Bones Segmentation

Presenter: Jérôme Schmid [MIRALab - University of Geneva, Switzerland]

Segmentation is a very active field in medical image analysis and many challenges are still present. One of them is the fast segmentation of low resolution clinical images which present noise and artifacts. This demo will depict semi-automatic bone segmentation from medical datasets. Our approach is based on a formulation of the deformable models evolution. Meshes vertices are considered as lumped mass particles subjected to internal and external forces. External forces are based on image features while internal forces regularize the models' evolution. The demo shows how 4 bones (left/right femur and hip bone) are simultaneously segmented.

Interactive Collaborative Segmentation

Presenter: Niels Nijdam [MIRALab - University of Geneva, Switzerland]

Image segmentation is the most common image manipulation tools that are used to define regions of interest (ROI) for communication among the users. The ability to collaboratively segment medical images introduces new opportunities, such as multiple experts segmenting the images concurrently, and for teaching and training purpose. This demo presents a prototype collaborative Teleradiology application that enables multiple users to concurrently segment the same image over a network. Our preliminary results suggest that such a system can improve communication and collaboration in telemedicine applications. The chosen segmentation approach is explained more in details in the other demo from the University of Geneva.

Ex vivo biomechanical testing of the knee joint

Presenter: Caroline Öhman [Laboratorio di Tecnologia Medica - Istituto Ortopedico Rizzoli, Italy]

The knee joint has been and is currently studied from a biomechanical viewpoint. The study of the joint degrees of freedom and of the forces/moments loading the joint is aimed at determining the function of a healthy or injured joint.

Ex vivo testing offers the chance to study deeply the knee joint biomechanics.

Using an experimental set-up, it is possible to replicate clinical experiments in order to determine the stiffness of the joint when intact and after rupture of ligaments. For example, clinically relevant measurements of stiffness, replicating either drawer tests or rotation, can be made.

Experimental ex vivo methods are being performed in our laboratory to determine the natural biomechanics of the knee, with two main outcomes: firstly, that of determining the biomechanical behaviour of healthy knees (which can then be used as a control to evaluate knee replacement devices, surgical repair, or the contribution of different components to physiological biomechanics); secondly, to provide data useful for development of knee joint models.

Studying Joints with MRI

Presenters: Bailiang Chen [University College London, UK]

Joints play an important role in controlling the body’s locomotion. To achieve better understanding of how they function, it is critical to know the anatomical structures and their kinematics more accurately. In the 3D anatomical human project, UCL is responsible for establishing protocols to acquire static and dynamic MRI data with appropriate clinical protocols as the basis for the information extraction and modelling in the further successive steps of the project. In this demonstration, the various imaging protocols on clinical scanners will be shown that have been established and tuned for acquiring high resolution static MRI images of lower limb and dynamic MRI studies of knee motion. In addition, Diffusion Tensor Imaging (DTI) of tibia and other bones and tissues is also been introduced to show a long-term aim of our research which is accessing the fine structure of such tissues. Innovative pulse sequences of musculoskeletal systems are presented and the concept of multi-scale imaging introduced. The demonstration shows the basic principles of MRI, as applied in particular to the domain of the image acquisitions required by the 3DAH project, and introduces new methods appropriate for the detailed analysis of the fine structure of tissues such as bone and cartilage in joints.

Extracting appearance information from medical images

Presenter: François Chung [Asclepios - INRIA, France]

Through small demos, methods to extract information from medical images are presented. A clustering of intensity profiles around a region of interest is performed on both synthetic and medical image. A spatial smoothing taking into account connectivity between profiles is depicted as well as the difficult task of pointwise correspondence, in our case by registering a current mesh onto a reference mesh. Finally, an appearance prior is presented and first results of its use for model based segmentation are shown.

Interactive finite element modeling of the knee joint

Presenter: Tobias Heimann [Asclepios - INRIA, France]

This demo features an interactive 3D simulation of a subject-specific knee joint. Ligaments and menisci are modeled using anisotropic finite element models that realistically capture the specific tissue properties. The user can interactively apply forces to the simulation, e.g. to flex the knee. Strain levels of the modeled ligaments are continuously visualized throughout the simulation, which allows to judge if the imposed movement is within the physiological range. The demo is implemented using the open source framework SOFA for medical simulations (www.sofa-framework.org).

Collaborative 3DAH: a framework for knowledge sharing

Presenters: Ioana Ciuciu and Han Kang [STARLab - Vrije Universiteit Brussel, Belgium]

The 3DAH content management system is intended to provide a framework for knowledge management integration inside the 3DAH project for the purpose of knowledge sharing. This is a distributed and collaborative tool which collects different parts of knowledge from domain-experts (e.g. physicians, researchers) in the form of templates which can be submitted online. Templates provide elementary data to fill the ontology framework by interacting with STARLab’s DOGMA server. We demonstrate the annotation of anatomical data by means of an ontological approach. Following this approach, content is classified and structured in order to create links between data with the purpose of information retrieval.

Towards a 5 d.o.f. surfacic driven knee

Presenter: Cédric Schwartz [SMI - Aalborg University, Denmark]

In most musculoskeletal models of the lower limb, the knee joint is modeled as a simple hinge joint. This 1 d.o.f. model cannot reflect the real complexity of the knee motion and in particular the screw-home-mechanism (internal/external rotation) . Moreover a simple hinge joint usually leads to bone collisions.

The knee stability is realized thanks to active (muscles) and passive (ligaments, menisci, articular surface shapes) components. All of these tissues should be taken into account in order to develop an accurate model of the knee. We here present a first step to such a model where the relative position of the tibia and the femur is evaluate by the optimization of the coherence of the surfaces in contact. In practice, the model is driven by the flexion/extension angle of the leg, which can be easily and quite accurately obtained with MOCAP systems.

This model aimed to respond to the objective of the 3DAnatomicalHuman EU Marie Curie Research Training Network: « developing realistic functional three-dimensional models for the human musculoskeletal system, the methodology being demonstrated on the lower limb ». Moreover, this model based on local surfacic information of the joint may provide patient specific information. The presented model is developed in SMI at Aalborg University (Denmark) thanks to MRI acquisitions provided by UCL (UK) and segmented by INRIA (France).

To create and execute neuromuscular models and simulations

Presenters: Nicolas Pronost and Anders Sandholm [VRLab - EPFL, Switzerland]

This demo will present models and features that will be introduced during the tutorial on Sunday evening. Advanced methods on data setup will be proposed as well as visualization and analysis modules within OpenSim. You will have the opportunity to fully perform predefined simulations but also to discuss about your own.

A Bolder Vision for the Exploration of Medical Datasets

Presenter: José Antonio Iglesias Guitian [Visual Computing group - CRS4, Italy]

During real-time medical data exploration using volume rendering, it is often difficult to enhance a particular region of interest without losing context information. In the demo session, we will present a new illustrative technique for focusing on a user-driven region of interest while preserving context information [1]. Furthermore, we will show latest advances of integrate variation of these ideas in our light field workstation [2] supporting the rendering of massive volumetric datasets [3].

[1] Luo, Y., Iglesias Guitián, J.A., Gobbetti, E., Marton, F.: Context preserving focal probes for exploration of volumetric medical datasets. In: Second 3D Physiological Human Workshop. (November 2009)
[2] Agus, M., Gobbetti, E., Iglesias Guitián, J.A., Marton, F., Pintore, G.: Gpu accelerated direct volume rendering on an interactive light field display. Computer Graphics Forum 27(3) (2008) 231-240. Proc. Eurographics 2008.
[3] Gobbetti, E., Marton, F., Iglesias Guitián, J.A. : A single-pass GPU ray casting framework for interactive out-of-core rendering of massive volumetric datasets. The Visual Computer 24(7-9) (2008) 797-806 Proc. CGI 2008.

<Back >